Chokes are an integral part of production systems and are crucial surface equipment that faces rough conditions such as high-pressure drops and erosion due to solids. Predicting choke health is usually achieved by analyzing the relationship of choke size, pressure, and flow rate. In large-scale fields, this process requires extensive-time and effort using the conventional techniques. This paper presents a real-time proactive approach to detect choke wear utilizing production data integrated with AI analytics.
Flowing parameters data were collected for more than 30 gas wells. These wells are producing gas with slight solids production from a high-pressure high-temperature field. In addition, these wells are equipped with a multi-stage choke system. The approach of determining choke wear relies on training the AI model on a dataset constructed by comparison of the choke valve rate of change with respect to a smoother slope of the production rate. If the rate of change is not within a tolerated range of divergence, an abnormal choke behavior is detected.
The data set was divided into 70% for training and 30% for testing. Artificial Neural Network (ANN) was trained on data that has the following inputs: gas specific gravity, upstream & downstream pressure and temperature, and choke size. This ANN model achieved a correlation coefficient above 0.9 with an excellent prediction on the data points exhibiting normal or abnormal choke behaviors. Piloting this application on large fields, where manual analysis is often impractical, saves a substantial man-hour and generates significant cost-avoidance. Areas for improvement in such an application depends on equipping the ANN network with long-term production profile prediction abilities, such as water production, and this analysis relies on having an accurate reading from the venturi meters, which is often the case in single-phase flow. The application of this AI-driven analytics provides tremendous improvement for remote offshore production operations surveillance.
The novel approach presented in this paper capitalizes on the AI analytics for estimating proactively detecting choke health conditions. The advantages of such a model are that it harnesses AI analytics to help operators improve asset integrity and production monitoring compliance. In addition, this approach can be expanded to estimate sand production as choke wear is a strong function of sand production.